ACCIDENTALLY INFORMED: INCIDENTAL NEWS EXPOSURE ON THE WORLD WIDE WEB By David Tewksbury, Andrew J. Weaver, and Brett D. Maddex An important element of news delivery on the World Wide YJeb today is the near ubiquity of breaking news headlines. What used to be called search engines (e.g.. Yahoo! and Lycos) are noiv "portals" or "hubs, popular services that use news, weather, and other content features to extend the time users spend on the sites. Traditional models of news dissemination in the mass media often assume some level of intention behind tnost news exposure. The prevalence of news on the disparate corners of the Web provides opportunitiesfor people to encounter current affairs information in an incidental fashion, a byproduct of their other online activities. This study uses survey data from 1996 and 1998 to test whether accidental exposure to news on the Web is positively associated with awareness of current affairs information. The results indicate that incidental online news exposure was unrelated to knowledge in 1996 but acted as a positive predictor in 1998. Many political theorists believe an informed public is a necessary Introduction ingredient for a healthy democracy,' Citizens are expected to be aware of important issues and to provide feedback to the political system,- As the primary information source for many people, mass mediated news is a key element in any contemporary political system,' As a consequence, evaluating the effectiveness of the media in performing their role is an important topic for scholars. There is a great deal of research on the traditional media and their effectiveness in transmitting information,^ However, thefielddoes not yet have a good grasp of the potential effects of the newest major medium, the Intemet, Researchers in this area are asking whether the Intemet provides a unique avenue for citizens to acquire knowledge about current events, Newsgathering has typically been seen as a purposeful, directed activity, Reading newspapers, listening to news on the radio, and , watching television news programs are normally seen as the result of conscious choices,^ In the traditional media, news is relatively segregated from other content. Audiences can quite easily turn on the television, for example, without coming across news stories. In contrast, David Tewksbury is an assistant professor in the Department of Speech Communication IbMC Quarterly and the Department ofPolitical Science at the University ofIllinois at Urbana-Champaign. Vo/,7S,;\o.,5 Andrew J. Weaver is a graduate student in the Department of Speech Communication at Autumn 2001 the University ofUlinois at Urbana-Champaign. Brett D. Maddex is a graduate student ©2ooi.if;-\/c in the Department of Communication at the University of Colorado, Boulder. The authors wish to thank Seth Finn for helpful commenta. 533 many of the most popular sites on the World Wide Web have integrated quite diverse areas of content on centralized services and pages. The Web may be unique in its ability to provide a typical user with an array of information choices that extend far beyond what he or she intentionally seeks. News headlines are an almost constant feature of the most frequently visited sites on the Web, and there is some evidence that people encounter current affairs information when they had not been actively seeking it.* Encounters of this sort may be called incidental exposure, and this may be an important contemporary avenue for citizen acquisition of current affairs information. Learningfrom the Traditional Media. Most views of learning from the media conceptualize it as an active process in which people are motivated to seek out and retain information about a specific subject.^ Often, citizens will turn to the media to learn about public affairs and politics and a host of other issues that concern them.* For example, newspaper readers commonly cite a need to keep up with the news as a reason for reading, and these people typically score higher on tests of issue knowledge.' These findings are quite robust. Even when researchers account for such demographic variables as age, education level, and income, newspaper reading is still a significant predictor of issue knowledge.'« Of course, not all knowledge gained from the media comes from active learning. Even those with little or no interest in public affairs can learn from the media." Passive learning, by definition, occurs in the absence of a motivation to become informed.^^ All that is required for information gain to occur is some minimal level of attention to the media in question. For example, people who watch the evening news regularly may learn about issues aired on the news, even when they are not interested in following those issues.'^ It seems clear that active and incidental learning are not mutually exclusive phenomena. Rather, there are undoubtedly a number of factors that can influence the blend of these two modes in practice. One such element may be the frequency with which people are exposed to specific news items. Researchers have found that news viewers have some difficulty in recalling information the day after it was presented on television news programs.'^ What is more, many viewers cannot recall any of the specific stories presented in a news show they have just watched.'^ Thus, there is evidence that exposure to one story presented a single time may not necessarily lead to learning of any significance. Important events and issues often receive extensive coverage over time and across news outlets, however. This repetition may be central to audience retention of the information.'* When the media present the same or similar stories over a period of time, they are giving the audience a chance to mentally rehearse the information. Even within television news broadcasts, viewers are often teased with major headlines before stories are shown. Such rehearsal allows individuals to retain the information, even in cases of passive learning.'^ News channel factors also can influence news processing style. Researchers have found that broadcast television news and newspaper use can vary in the types of learning that result. Individuals who are actively seeking information tend to go to newspapers for that information, as newspapers offer more depth and easier access to particular stories."* Partly as a result of this active information seeking, newspaper readers tend to be more informed than those who rely on other forms of media for news.'' Television viewers, on the other hand, may tend towards more passive learning.^*^ A television news viewer may be interested in only sports and weather, but the viewer will often sit through the opening news stories to get to those segments. Again, this kind of exposure can lead to learning, even without direct intention. The television format does not allow the same kind of depth and thoughtful processing (e.g., rereading, time for elaboration, etc.) that the print media allow. However, television news is more widely accessible, and the use of graphics and visuals leads to a more memorable and more digestible presentation.^^ Thus, television \ iewers are able to acquire information at relatively low levels of attention. Clearly there is a difference between print and television news, but they contribute in complementary ways to public information gain. Though newspaper use typifies active learning, relatively passive learning may still occur. As a reader flips through the paper, headlines may flash into consciousness whether the reader is interested or not. Similarly, while television is largely associated with passive learning, active learning can still occur through television viewing. For example, if a viewer tunes in to see a particular local news story, that viewer is actively gathering information on that issue. Furthermore, a combination of the different media can lead to a stronger learning effect along the lines of repetitive learning described above. If people see the same story on the evening news that they read about in the morning paper, the repetition will form stronger connections in memory. Learning from the World Wide Web. Comparisons among print, television, and radio as sources for news learning have been going on for decades.^^ Recently, the Internet has been added to the mix. Early research on computer-mediated news indicated that knowledge gain from use of that medium may be equal to that obtained from newspaper reading.^ This is not surprising given that early computer-mediated news was very similar to printed copy. However, things have changed since the early 1990s; the World Wide Web of today is much more interactive and offers a wider range of content and formats. Web users can now view streanüng audio and video, and they can follow \ii\ks to a wide array of background information. They can even customize their online experience to fit their personal needs.^* Given the wealth of information available on the Web, it would seem that at the very least this medium would be an arena for active, goaldirected learning.^ The layout of the Web is ideal for searching for specific information. Many of the most popular sites are search engines that direct people to the information they want. Indeed, almost half of the people orüine are attracted to Web-based news because of the ability to search for news on a specific topic that interests them.^^ However, while most media use is goal directed, not all goals involve the acquisition of information. In order to make a distinction between intentional and unintentional learning on the Web, it is imporAcciDCNTAUY INFORMED 535 tant to look at the different ways people use the medium. The uses and gratifications approach posits that people use media to fulfill certain goals, and different uses can lead to different effects.^ This is useful in looking at the varieties of learning that come from Web use because, again, people have different goals when going orüine.^ For example, a primary use of the Internet involves information seeking.^ This is in line with research that has found the Internet is an outlet for active learning.* However, many Internet users also find pleasure in simply surfing, following links to xmknown destinations.^' Others núght use the Web to chat or play games. These are situations in which any current events learning that occurs is likely to be less than active. There is some evidence that there is less current affairs learning from Web use than there is from the traditional media. For example, when a comparison was made between a print newspaper and the electronic version of that newspaper, readers of the print version scored better in a test of public affairs knowledge.''^ Moreover, there is a question of whether unintentional learning could occur on the Web as it does in the traditional media. It may be that relatively few Web users go online to browse through news sites as one might browse through a newspaper. Most Internet use is goal directed,'^ and acquiring current affairs information is rarely a primary goal.^ Therefore, it would seem that incidental learning through use of the Web would not be sigrüfícant. However, there are reasons to believe that nonpurposive news contact does occur on the World Wide Web, and more important, this incidental nwvs exposure could positively influence leanüng about current events. What may be happening for Web users is not so much passive learning of the variety Krugman^ and Zukin and Snyder* have described but learrüng that proceeds from spontaneously generated interest. That is, interest may be piqued by a headline or by the mere presence of a list of news items. Graber has demonstrated that audience interests are an important component of the decision to read specific newspaper articles.^^ We believe that the Web provides an important new route by which people are exposed to news. We believe audiences may see a number of news headlines in the course of their Web use. On occasion, their interest is aroused long enough for them to register a headline and perhaps click cind read the accompanying story. In this way, exposure to news occurs when people may have been using the Web for reasons enfirely unrelated to informafion seeking. Two factors are central to current patterns of incidental online news exposure: the move by many Web sites from being simple search engines to acting as iriformation services and the substantial popularity of these first-stop or portal sites. The portal sector of the Internet industry believes that transforming search engines and first-stop Web sites to information and/or personalized services is not orUy advantageous but is also necessary to remain competitive.^ On almost all portal Web sites (e.g.. Yahoo!, Lycos, and Excite), news of the day and/or breaking news stories are displayed on the welcome page, the starting point for most Web activity. Users can elect to have news from additiorul categories, such as international news, political news, or sports, displayed as well. 536 ¡OUKNALBM&MASSCOUUBJMCJinONQUMtaB While many Web sites are doing some version of this "infomating," the portal sites have made this a mainstay of their business strategy.^ This is significant because a number of these portal or first-stop services are among the most popular sites on the Web. In fact, of the seven most frequently visited Internet properties for March 2001, four were primarily portal sites (e.g., Lycos) and two featured sites designed like portals that carried news of the day on the front page (e.g., MSN).*" These top seven sites were visited by 18 percent to 65 percent of the people who used the Web that month, and people spent more than three times as much time on portal sites as they did on news sites.*' The result of these developments is that users are increasingly likely to encounter news items, even when searching the Web for spedfic non-news information. If a similar level of news saturation was present in one's offline activities, it would be akin to coming across news headlines posted on the front of a telephone book, on the front door of the bookstore, or at the beginning of telephone conversations. It would almost be as though television and radio audiences received news headlines every time they tuned their receivers in search of content. We may be overstating the case a bit, but the point is that for many people news seeking is not a core media habit. We believe these people may encounter news more frequently on the Web than they would offline. The prevalence of news on the Web could increase the learning of current events issues through the repetitive nature of exposure. Repetition of news stories in the traditional media increases knowledge of those issues.*^ Presumably, this same effect occurs when people are exposed to stories many times on the Web. Moreover, people typically use the Web in conjimction with other media. A person may read the moming newspaper, use the Web at work, and then watch the local television news in the evening. This allows for repetition of the stories across media. Because the Internet is often an addition to rather than a replacement of traditional media, we believe that Web use will increase incidental learning about current events in part because of the complementary nature of contemporary media.*^ Of course, before we can get to tin increase in learning coming from an increase in exposure to news headlines, it is necessary to look at whether incidental exposure to news headlines is a function of Web use. We expect that the more time people spend on the Web, the more they should come across news. This should be the case even if they are not specifically looking for news items. So, the first research hypothesis 'states, HI: The frequency of Web use will be positively associated with incidental exposure to news on the Web. Following the uses and gratifications approach, the goals people have for surfing the Web could impact incidental exposure. Clearly, if people are looking for particular current events information, they are more likely to come across other news items. However, we are also interested in how other goals relate to incidental exposure. 537 RQl : Do different uses of the Web result in differences in incidental exposure to news? Assuming that Web users are incidentally exposed to news headlines, our primary concern is whether this additional exposure to news has an effect. Because of the repetition of news stories that Web users experience, and because relatively passive learning has been demonstrated in other media, we believe we will see incidental learning in this context, H2: Incidental exposure to news on the Web will be positively associated with knowledge of current events. Method Data useful for testing our expectations were available from the Pew Research Center for the People and the Press, At irregular intervals the Pew Center conducts national surveys of media and communication technology use by the American public. Surveys relevant to the present research questions were taken in Autumn (21 through 31 October) 1996, Spring (24 April through 11 May) 1998, and Autumn (26 October through 1 December) 1998, The first and third surveys were specifically focused on Intemet use, while the second was more broadly concerned with news media use. Identical or nearly identical question wording was used for a number of the variables of interest. However, there are cases in which only a subset of the items is available in any one survey. Taken together, however, they were adequate for the task at hand. The data allowed us to examine both our central research hypothesis and to examine the potential antecedents of incidental news exposure. Perhaps the biggest challenge with testing our expectations with survey data was the measurement of the primary independent variable, incidental exposure to news on the Web. Included in all three surveys were measures of current affairs knowledge and incidental exposure to news while online,'" As one would expect, there is no ideal way to measure the latter concept in a survey context. The Pew data featured a relatively straight-forward question (see below), but it is likely that a retrospective assessment of one's own experiences may suffer from some level of unreliability. As an alternate approach, one could approximate the likelihood of incidental news exposure with measures of respondents' overall time online minus their time online spent seeking news. That is, the conceptual definition of incidental news exposure is encountering current affairs information while performing some other task. The likelihood of such exposure occurring should increase with use of the Web with a goal unrelated to news seeking. Where possible, we performed parallel tests of our incidental exposure hypothesis. We used a dichotomous self-report of such occurrences (a relatively direct measure) in some analyses and overall time online, controlling for online news seeking (a relatively indirect measure), in others. While both measures undoubtedly carry some level of error in tapping the concept that interests us, we hoped that a dualpronged approach would help us overcome some of that error. In all tests 538 of our expectations, we attempted to control for factors that may have produced a spurious relationship between our independent variables and current affairs knowledge. Thus, our multivariate analyses featured measures of respondents' interest in politics and political news, their general news exposure patterns, their frequency of online news seeking, and demographic variables (e.g,, education, gender, and income) that may be associated with political knowledge acquisition.^' Samples. The 1996 survey was a telephone re-interview of 1,003 adult respondents who had reported Internet use in previous Pew Center studies and a small random sample of those who had reported ownership of a computer but no Internet use. Both 1998 surveys were conducted for the Pew Center by Princeton Survey Research Associates. The Spring data were from a national random-digit (RD) sample of 3,002 adults. Only a subset of respondents («=1,192) reported use of the Internet, of course. All analyses of these data were limited to that subset. The Autumn data were drawn from a national RD telephone survey of 3,184 adults. This dataset included an oversample of Internet users. There are 1,993 respondents available for the present analyses/** Current Affairs Knowledge. Each of the Pew Center surveys contained a handful of current affairs questions. Question wording for these and other measures is given in the Appendix. The 1996 survey measured respondents' knowledge of two elements of the presidential campaign of that year, candidates' positions on tax cuts, and the owner of the phrase "bridge to the future." Another question assessed awareness of which political party controlled the U.S. House of Representatives. Responses to these items were recoded (correct = 1, all others = 0) and sununed to form an index of public affairs awareness (M = 2.16; SD = .92). There were three questions available in the Spring 1998 survey, but two of them had to be discarded. The first was an item that asked respondents whether high cholesterol is generally considered to be bad or good for one's health. Ninety-six percent of respondents answered this correctly, so it was not particularly diagnostic of differences in knowledge level. The other item concerned the direction of weather pattern movement in the United States, This question did not appear to tap the desired concept in this case. The question retained for analysis asked for the identity of the Speaker of the U.S. House of Representatives. Sixty-four percent of respondents correctly answered this nominal measure of knowledge (correct = 1, all others = 0). Finally, the Autumn 1998 survey included three relevant questions: Items about the federal government's antitrust case against Microsoft, the identity of the party in control of the U.S. House of Representatives, and the names of countries who had recently exploded nuclear devices. Correct responses to the third item were India, Pakistan, or both. Respondents who answered with one were scored with a .5, those who gave both received a 1, and those who could provide neither were given a 0. This item was added to the others (both recoded as 1 = correct and 0 = all other responses). The mean score on this index was 1.77 (SD = .98). The survey also included an item about the meaning of the "org" suffix in Internet addresses. However, this question did not tap current affairs knowledge so it was discarded. 539 Incidental Online News Exposure. Unintentional exposure to Web-based news was assessed with a simple yes/no question in all three surveys (see Appendix for wording). Responses were recoded (yes = 1, no = 0). In each of the surveys, about half of the sampled Internet usen responded in the affirmative (53% in 1996; 55% and 49% in Spring and Autumn 1998, respectively). Frequency of General and News-Oriented Use. Overall online exposure and use with the goal of news reading were assessed with parallel questions in the 1996 and Autumn 1998 surveys (see Appendix for wording). Only the latter was measured in the Spring 1998 study. All responses were given on a 6-point scale (every day, 3 to 5 days per week, 1 or 2 days per week, once every few weeks, less often, or never). These responses were recoded to more closely resemble an interval level scale by assigning the values of every day = 7,3 to 5 days per week = 4,1 or 2 days per week = 1.5, once every few weeks = .5, less often = .05, and never = 0. The mean response for overall online use frequency in both surveys was almost 4 days (M = 3.58 in 1996,3.73 in Autumn 1998; SDs = 2.46 and 2.54, respectively). Naturally, the mean responses for online news use were subsets of these levels (Ms = 1.50 and 1.52, SDs = 2.10 and 2.27, respectively; M = 2.37 and SD = 2.61 in the Spring 1998 survey). Online Activities. In the 1996 and Autumn 1998 surveys, a battery of questioris measured the genercd character of respondents' activities on the Web. Four items were included in both studies: commimicating with others, obtaining financial information or trading securities, getting news, and obtaining entertainment-oriented content (see Appendix for wording).*^ The 1996 survey also included a measure of going online to obtain work-related information. This question was included in the Autumn 1998 survey, but it was administered to only half of the respondents. Its inclusion in those analyses would have dramatically reduced the available sample size, and so it was discarded. Afinalquestion asked of all respondents in Autumn 1998 dealt with going online to get health or medical information. Responses to all of the items were given on a 6-point frequency scale, subsequently recoded as described above. Information seeking for work turned out to be the most frequent activity in 1996 (M = 1.86 days per week, SD = 2.25) and obtaining entertainmentrelated information was the least frequent (M = .84, SD = 1.40). News seeking, a close second in 1996, was the most frequent activity in Autumn 1998 (M = 1.45, SD = 2.17) and obtaining health-related information was the leastfirequent(M = .50, SD = 1.22). interest in Politics and News. The surveys did not contain a consistent measure of interest in politics. An item in the 1996 study asked respondents about their level of interest in politics, and a question in Spring 1998 asked how much they generally kept up with the news. An item Üiat used the traditional National Election Studies assessment of following politics appeared in the Autumn 1998 study. Responses to all three questions were given on 4-point ordirud scales (scored 1 to 4; see Appendix for wording) recoded so that higher scores indicated greater interest in news and politics. For the purposes of the analyses reported here, they were taken to approximate interval-level scales. The mean for 540 the 1996 question was 1.94 (SD = .81). For the Spring 1998 item it was 3.36 (SD = .75), and in the Autumn it was 3.24 (SD = .94). News Exposure. Participants' newspaper and television news exposure were assessed with two-part questions. The first was a filter that asked whether respondents had read the previous day's paper or watched television news the day before the study. Those answering yes were probed for the number of minutes they spent reading or watching. The duration values were used in the present analyses with respondents reporting no exposure the previous day receiving a value of 0. Radio news exposiire was assessed with one question with no-listening given as one of the response options (recoded to zero here). All three duration questior\s were recoded from an ordinal scale to a form more closely approximating an interval measure (less than 15 minutes = 7.5, 15-29 minutes = 22,30-59 minutes = 44, and 1 hour or more = 60). Demographics. Included in each of the surveys were gender (coded female = 1, male = 0 here) and age. The proportion of females in the samples increased from 39% in 1996 to 46% in Spring 1998 and 50% later that year. Mean ages were 38 in 1996 and Spring 1998 and 39 in Autumn 1998. Also included in all three surveys were measures of education and income (see Appendix for question wording). As is usually the case, the income variable had a substantial portion of missing data, presumably due to refusals. The mean education level in 1996 was college graduate; in the later surveys it was "some college, no 4-year degree. " The median income level in 1996 and Autumn 1998 (the Internet user samples) was $50,000 to under $75,000. In Spring 1998 it was $40,000 to under $50,000. Predicting Accidental Exposure. On an operational level, the first hypothesis states that overall exposure to the Web should be positively associated with self-reports of incidental exposure to news there. Data for testing that expectation were available in the 1996 survey and the Autumn 1998 study. On a bivariate level, there is rrüxed support for the hypothesis. A measure of nominal-by-interval association (TI) shows no real relationship between online use frequency and the nomii\al measure of incidental exposure in 1996 (r|= .03). However, the association in Autumn 1998 is significant (i\ = .16).** In an effort to remove the effects of tíürd factors, we performed logistic regressions in which incidental exposure was the dependent variable and demographic factors and the .Web activity items were independent variables. The results are showm in Table 1. The logistic regression coefficients for overall Web use support the bivariate findii»gs. In 1996, there is no meaningful association but in 1998 there is. In summary then, there is orüy mixed support for the first hypothesis. The data in Table 1 address our research question, as well. The issue here is whether we can identify the Web activities associated with self-reported incidental news exf>osure. In both cases, we find that, controlling for the overall number of days respondents go online and for thefrequencyof selected other activities, seeking current events ir\formation increases the likelihood of incidental exposure. In the 1996 study. 541 TABLE 1 Likelihood of Reporting Accidental News Exposure Autumn 1996 Autumn 1998 Demographic Variables Female -.30 (.17) -.14 (.11) Age -.03 (.Ol)»" -.02 (.00)* Education -.00 (.07) .07 (.04) Income .06 (.05) .02 (.03) Overall Online Frequency -.02 (.04) .09 (.02)*»» Online Activities Communicate with Others .08 (.05) -.00 (.03) Get Financial Information .07 (.04) -.02 (.03) Get Information for Job -.12 (.04)»» Get Current Events News .16 (.04)»» .19 (.03)»»» Get Entertainment Information ,02 (.06) .04 (.03) Get Health/Medical Information Constant Model Chi-Square N — 1.04 (.45)» 48,74»»» 731 .05 (.05) -.21 (.26) 128.38» 1,604 Note: Dependent variable is self-reported occurrence of incidental online news exptosure (yes = 1, no = 0). Female is coded 0 = male, 1 = female. Cell entries are logistic regression coefficients. Entries in pctrentheses are standard errors. using the Internet to obtain infonñation for one's job was also a predictor, but this activity tended to reduce the likelihood of accidental exposure. While a number of tiie other activities exhibit bivariate associations with incidental exposure (e.g., in the 1996 data, commimicating with others, getting entertainment information, and getting financial information were positively correlated with incidental exposure, all rs = .11, ps < .01), these relatior\ships washed out in the regression analyses. In summary, the situation is much as one might expect: Those who tend to look for news orrline are the ones who tend to come across it by accident as well. 542 TABLE 2 Antecedents of Current Affairs Knowledge Autumn 1996 Spring 1998 Autunm 1998 Demographic Variables Female -.29 (.Oó)»" -.27 (.06)"* -.52 (.20)» -.42 (.04)»« -.43 (.04)"* Age .00 (.00) .00 (.00) .01 (.01) .01 (.OO)"» .01 (.00)»»» Education .11 (.03)*»» .11 (.02)»" .25 (.07)»*» .15 (.02)»»» .16 (.02)»»» Income .05 (.02)*» .05 (.02)" .09 (.05) .04 (.01)»» .04 (.01)»»» .32 (.04)»»» .35 (.04)»» ,38 (.15)» .28 (.02)»»» .28 (.03)»»» Newspaper -.00 (.00) -.00 (.00) .01 (.01) .00 (.00) ,00 (.00) Radio News .00 (.00) .00 (.00) .00 (.00) .00 (.00) .00 (.00)»» Television News .00 (.00) .00 (.00) -.00 (.00) -.00 (.00) -.00 (.00) Online News .03 (.02)» .01 (.01) -.05 (.04) -.02 (.01) .03 (.01)»» Interest in politics/news News Media Expœure Overall Online Frequency Incidental Exposure to News .04 (.01)»»» -.03 (.01)* — .04 (.06) .38 (.22) — Constant .64 (.17)» .46 (.16)»» -2.71 (.59)»»» -.50 (,11)»»» R-squared / Chi-square N .20»»» 733 .22»»» 863 59.76»»» 516 .36»»» 1,591 — ,14 (.04)»» -.52 (.11)» .35»»» 1,585 Note: Cell entries for Autumn 1998 and 1996 are unstandardized OLS regression coefficients. Cell entries for Spring 1998 are logistic regression coefficients. All entries in parentheses are standard errors. . Interest in politics/news question wording varies by survey. See appendix for wording. Female is coded 0 = male, 1 = female; incidental exposure to news online is coded 0 = no, 1 = yes. It appears that accidental news exposure might be happening, in part, among those already oriented toward news. This highlights the necessity for our analyses of the second hypothesis to include adequate controls for news consumption behaviors. Finally, of the demographic factors, only age exerted a significant independent influence such AcanomuybiFaaiED 543 that older respondents were less likely to report incidental news exposure. Incidental Exposure and Current Affairs Knowledge. The second hypothesis states that incidental exposure to online news will have a positive association with current affairs knowledge. The bivariate relationships between current affairs knowledge and self-reports of incidental exposure show an inconsistent pattern. In Spring 1996, the two variables appeared unrelated (r\ = ,04), However, in both of the 1998 studies, there is a significant positive relationship (x^(l, 578) = 6.97,p< .01 in Spring 1998; TI = .14 in Autumn 1998)."' Table 2 presents the results of our multivariate analyses. The Autumn 1996 and 1998 sets of columns present the results of OLS regressions of current affairs knowledge indices on demographic and other control factors. Because we had two indicators of incidental exposure (overall frequency of online use and self-reports of exposure), there are two equations for each survey. The Spring 1998 study featured only one usable knowledge question, so the third column reports the results of a logistic regression of that dichotomous item on the antecedents. Also, because an overall measure of Web use frequency was not available in that survey, only one set of results is reported. Across the board, the measures of demographics and a general interest in politics were significant predictors of current affairs awareness. Males were more likely to correctly answer the questions, as were those with higher education and income. In Autumn 1998, age also emerged as a positive predictor of knowledge. Much as one would expect, interest in news and politics—as it was variously measured— was a consistent positive predictor of knowledge. Surprisingly, exposure to both traditional and orüine news outlets exhibited little effect in these data. They were all moderately associated with current affairs knowledge on a bivariate level (rs and ris typically in the .05 to ,15 range in all three surveys); however, their effects were washed out by the other variables in the equations. The key test for the second hypothesis was whether the indirect and direct indicators of incidental exposure would predict current affairs knowledge once other factors were controlled. There are some interesting results here. First, it is clear that the 1996 and 1998 data behave very differently from one another. In 1996, the direct measure of incidental exposure failed to exert an influence on knowledge. The coefficient for overall frequency of use is significant, but its sign is opposite what was expected. Thus, with frequency of orüine news use controlled, additional online activity was associated with lower current affairs knowledge. In Spring 1998, with only the self-report of incidental exposure and one knowledge item available, the association between them is marginally significant (p=.O8). This is closer to what was expected. The final survey exhibits relationships that fall in line with expectations. Both overall time online and self-reported incidental exposure to news emerge as positive predictors of current affairs knowledge. To be sure, the effect they are exerting is not large (the standardized coefficient for the self-report measure is .07), but it is equivalent to that observed for intentional online news seeking (its beta is also .07). Thus, 544 above and beyond the effects of relevant demographic variables, other news media exposure, and a general tendency to follow politics, our two indicators of incidental exposure to online news exert a significant, though small, positive influence on current affairs knowledge in the late 1998 sample. The results of our analyses offer some evidence that incidental exposure to news while using the Web is a phenomenon worthy of further investigation. As expected, the more frequently people went online in 1998, the more likely they were to report unintentional news exposure. Much as one might expect, the tendency to report this sort of exposure was positively associated with users' news-seeking activity. We had asked whether other Web use goals might emerge as correlates of incidental exposure, but none was a significant predictor in the multivariate tests. We found in these three sur\ eys that approximately half of all respondents reported incidental news exposure. Those who did tended to be a bit younger and more likely to go online for news gathering than were other users. More important, we find some evidence that unintentional exposure can lead to greater knowledge of current affairs. To be sure, the picture is mixed, but there are some reasons to conclude that this hypothesis has support. We did not find any level of association between self-reported incidental exposure and current affairs recall in 1996. Additionally, the indirect measure of incidental exposure exhibited a sigrüficant negative relationship in that study. In Spring 1998, the relatioriship was in the predicted direction but was only marginally sigruficant. The Autumn 1998 analyses revealed a consistent, positive association of incidental exposure (measured two ways) with recall. The task now lies in explaining this pattern. There are a number of possible explanations for the surprising findings in 1996 and the positive findings in late 1998. Thefirstpossibility is that the nature of the samples was meaningfully different in these surveys. The three surveys used here contained only people with Internet access, of course. It is possible that the population of Internet users changed sufficiently over the intervening two years for a different set of behaviors and effects to have been present. It is clear that the size of the Internet audience increased dramatically in that period, climbing from 23 percent of the American population to 41 percent and became demographically more like the general populace.™ As the samples became more representative of everyday Americans, it could be that their habits, in the aggregate, changed in the manner exhibited in these data. This may account for the null findings in 1996 and the positive results in 1998 for both of our hypotheses. A second possibility is that some of the pattern we see across these surveys is the result of measurement error. Certainly, the Spring 1998 data are suspect in this regard. There is only one measure of current affairs knowledge (a norrunal recall item), and there is only one indicator of incidental news exposure. Although the association between them was as predicted in both bivariate and mulfivariate settings, the signifi- Lftscusston 545 canee level of the latter was only marginal. The measurement error explanation may be able to account for part of the difference in findings between 1996 and late 1998, but this alone may not fully explain the disparity, A third explanation is that the online environment was very different in 1996 than it was in 1998, From this line of reasoning, the opportunity for incidental exposure was much greater in 1998 than two years earlier. There is some evidence in support of this explanation. We believe a substantial portion of incidental exposure occurs as people are going about their everyday activities on the Web, As the "portal" concept took off in 1997 and 1998,^' the inclusion of news headlines became an important content feature designed to keep Web users from straying too far away,''^ Thus, while many important news outlets (e,g,, CNN and the New York Times) had a substantial Web presence in 1996, users at the time were less likely to encounter news on other popular sites than they were two years later. The presence of an association between the incidental exposure measure and frequency of use in 1998 but not in 1996 offers some support for that explanation. Although people reported incidental news exposure just as frequently in 1996 as in 1998, it may be that respondents in 1996 were thinking of only occasional exposure and those two years later were thinking of frequent exposure. In other words, despite coming across news less frequently, people in 1996 were still able to answer the exposure question in the affirmative. Their relatively infrequent exposure meant that they were unlikely to incidentally acquire current affairs information in any meaningful sense. Our indirect measure of incidental exposure (i,e,, online use frequency, controlling for news exposure) also failed as a predictor of knowledge in 1996 because the paucity of news on non-news sites meant that a measurable acquisition of information was likely fo occur only through purposeful news seeking. As a result of these factors, we failed to find the expected link between knowledge and our independent variables in 1996, By 1998 the situation had changed, and our indicators of incidental exposure emerged as reliable, positive predictors. Among the limitations of this study is the lack of more recent data. The Pew Center omitted the incidental exposure question and overall use frequency measure in its next general Internet study,^' and we have not found a replacement source for these data, We also would have preferred better measures of current affairs knowledge. In the best of the available indices (Autumn 1998), there were only three questions. Also lacking is a more precise measure of accidental news exposure. Given the natural constraints of survey research, conducting an experiment ought to be a better approach. Future research may want to head in that direction. Another important limitation lies with the relatively small effect sizes here. The two indicators of incidental news exposure each accounted for only about 1 percent of the variance of the dependent variable in Autumn 1998, Thus, while the impact of incidental exposure is statistically significant, we do not have evidence that it is substantial. Of course, the less than ideal measurements of knowledge and incidental exposure may be partly responsible here. It may be that some level of random error has obscured the impact of incidental exposure. Unfortu- 546 nately, we cannot be sure. Finally, we would have liked a general measure of political knowledge, one that contains background rather than current affairs knowledge. Prior research has demonstrated that the knowledgeable are the ones most likely to acquire new information from the media.** We would certainly be interested in how that operated in the context of incidental exposure. In many ways, the idealized citizen is one who lives for the body politic. He or she continually monitors the information envirorunent, sifts through the available news, talks to family and friends, and provides input to the political system.^ It is unlikely that many such citizer\s exist in a modem democracy. Rather, there is likely great variance in political and informational involvement in any population. There are those who actively seek current affairs information, but a great many probably see it as merely a greater or lessor piece of their overall media diet. For this group of people, incidental exposure to the news may be an important avenue for acquiring information. Indeed, we found in late 1998 that 41 percent of Internet users who claim to follow govemment and public affairs "most of the time" reported incidental news exposure but among those who claim to follow it "hardly at all," 51 percent reported it (Í 11,167] = 2.35, p < .05). In other words, the World Wide Web of today may provide a public space where a broad cross section of the population encounters news not purposively but accidentally while going about daily business. Certainly this has been possible and maybe even frequent with the traditional media. However, the convergence of such a disparate range of activities on this one medium means that the Web provides the infrastructure for a much wider dissemination of breaking news headlines. For this reason, perhaps more than for many hyperbolic claims one hears, the Web may be a positive force in American politics. Appendix and Notes follow. 547 APPENDIX Question Wording Current Affairs Knowledge Autumn 1996 * As far as you know, which presidential candidate calls for a 15 percent across-the-bocird income tax cut and a 50 percent cut in the capital gains tax? Is it Bill Clinton, Bob Dole, or Ross Perot? * Do you happen to know which political party has a majority in the U.S. House of Representatives? » Do you happen to know which candidate has used the phrase "bridge to the future"? Is it Bill Clinton, Bob Dole, or Ross Perot? Spring 1998 Who is the Speaker of the U.S. House of Representatives? Autumn 1998 » Do you happen to know which computer software company is involved in an anti-trust dispute with the Justice Department? * Do you happen to know which political party has a majority in the U.S. House of Representatives? * Can you name any of the countries that recently exploded nuclear weapons? If respondent named one, interviewer probed once. Can you name any others? Responses were coded as India, Pakistan, some other country, or don't know/refused. Incidental Online News Exposure Overall online frequency Online news use 548 Autumn 1996 When you go on-line, are you ever exposed to news and information on current events, public issues, or politics when you may have been going on-line for a purpose other than to get the news? Spring & Autuumn 1998 When you go online, do you ever encounter or come across news and infonnation on current events, public issues, or politics when you may have tieen going online for a purpose other than to get the news? Autumn 1996 How often do you go on-line to (Intemet provider) every day, 3 to 5 days per week, 1 or 2 days per week, once every few weeks, less often, or never? Autumn 1998 How often do you go online... every day, 3 to 5 days per week, 1 or 2 days per week, once every few weeks, or less often? Autumn First, how often do you go on-line to get news and /oumwsMfirJMASS CcMMiNCQXw QuiuncMy Online Activities 1996 information on current events, public issues or politics, everyday.. .3-5 days per week, 1 or 2 days per week, once every few weeks, less often, or never? Spring & Autumn 1998 How frequently do you go online to get news... would you say every day, 3 to 5 days per week, 1 or 2 days per week, once every few weeks, or less often? Autumn 1996 & 1998 How often do you go on-line to (activity) everyday, 3-5 days per week, 1 or 2 days per week, once every few weeks, less often, or never? * Communicate with other people through online forums, discussion lists, or chat groups * Get financial information such as stock quotes or corporate information or to buy stocks or bonds? * Get news and information on current events, public issues or politics * Get information about hobbies, movies, restaurants or other entertairunent-related activities Autumn 1996 Autimm 1998 Interest in Politics/News Newspaper, Television, and Radio Exposure • Look for information for your work or job * Get health or medical information Autumn 1996 Generally speaking, how much interest wovild you say you have in politics: a great deal, a fair amount, only a little, or no interest at all? Spring 1998 How much do you enjoy keeping up with the news? A lot, some, not much, or not at all? Autumn 1998 Some people seem to follow what's going on in government and public affairs most of the time, whether there's an election or not. Others aren't that interested. Would you say you follow what's going on in government and public affairs most of the time, some of the time, only now and then, or hardly at all? Autumn 1996; Spring & Autumn 1998 * Did you get a chance to read a daily newspaper yesterday, or not? If yes: About how much time did you spend reading a daily newspaper yesterday? Responses categorized by interviewer as less than 15 min., 15-29 min., 30-59 min., 1 hour or more, or don't Appendix A cont. next page AoaoammbaoiiMCD 549 Appendix A cont. know/refused. * Did you watch the news or a news program on television yesterday, or not? If yes: About how much time did you spend watching the news or any news programs on TV yesterday? Responses recorded as above. * About how much time, if any, did you spend listening to any news on the radio yesterday, or didn't you happen to listen to the news on the radio yester day? Responses categorized by interviewer as less than 15 min., 15-29 min., 30-59 min., 1 hour or more, didn't listen, or don't know/refused. Educafion Autumn 1996; Spring & Autumn 1998 Income Autumn 1996; Spring & Autumn 1998 550 What is the last grade or class that you completed in school? Responses categorized as none, grade 1 - 8; high school incomplete (Grades 9-11); high school graduate (Grade 12 or GED certificate); business, technical or vocational school after high school; some college, no 4-year degree; college graduate (B.S., B.A., or other 4-year degree); post-graduate training or professional schooling after college; or don't know/ refused. . Last year, that is in (prior tax year), what was your total family income from all sources, before taxes? Just stop me when I get to the right category. Less than $10,000, $10,(KX) to under $20,000, $20,000 to imder $30,000, $30,000 to under $40,000, $40,000 to under $50,000, $50,000 to vmder $75,000, $75,000 to imder $100,000, $100,000 or more? NOTES 1. John Stuart Mill, "Of the Liberty of Thought and Discussion," in Utilitarianism, On Liberty, and Considerations ofRepresentative Government, ed. Harry Acton (London: J.M. Dent, [1859] 1972). 2. For a different perspective, see Michael Schudson, The Good Citizen: A History of American Civic Life (Cambridge, MA: Harvard Urüversity Press, 1998). 3. Harold D. Uisswell, "The Structure and Function of Commimication in Society," in Mass Communications, ed. Wilbur Schramm (Urbana, IL: Umversity of Illinois Press, 1960), 117-30. 4. For reviews see Doris Graber, Mass Media and American Politics, 5th ed. (Washington, DC: Congressional Quarterly Press, 1997); Steven H. Chaffee and Stacey F. Kanihan, "Learning About Politics from the Mass Media," Political Communication 14 (October-December 1997): 42130. 5. For examples see Charles Atkin, "Instrumental Utilities and Information Seeking," in New Models for Mass Communication Research, ed. Peter Clark (Beverly Hills, CA: Sage, 1973); Mark R. Levy and Sven Windahl, "Audience Activity and Gratifications: A Conceptual Clarification and Exploration," Communication Research 11 (January 1984): 5178. 6. Pew Research Center, 77K Internet News Audience Goes Ordinary: Online Newcomers More Middle-Brow, Less Work-Oriented (Washington, DC: Author, 1999). 7. Atkin, "Instrumental Utilities and Informafion Seeking"; Richard C. Vincent and Michael D. Basil, "College Students' News Gratifications, Media Use, and Current Events Knowledge," Journal of Broadcasting & Electronic Media 41 (summer 1997): 380-92. 8. Michael Gurevitch and Jay G. Blumler, "Political Commimication Systems and Democratic Values," in Democracy and the Mass Media, ed. Judith Lichtenberg (New York, NY: Cambridge University Press, 1990), 269-89; Lasswell, "The Structure and Function of Commurücation in Society," 117-30. 9. Dan Berkowitz and David Pritchard, "Political Knowledge and Communication Resources," Journalism Quarterly 66 (autumn 1989): 697701; Hugh M. Culbertson and Guido H. Stempel HI, "How Media Use and Reliance Affect Knowledge Level," Communication Research 13 (October 1986): 579-602. 10. John P. Robir\son and Mark R. Levy, "News Media Use and the Informed Public: A 199O's Update," Journal cf Communication 46 (spring 1996): 129-35. 11. Jay G. Blumler and E>erüs McQuail, Television and Politics: Its Uses and Influences (Chicago: University of Chicago, 1969). 12. Herbert E. Krugman and Eugene L. Hartley, "Passive Learning Through Television," Public Opinion Quarterly 34 (summer 1970), 184-90. 13. Cliff Zukin and Robin Snyder, "Passive Learning: When the Media Environment is the Message," Public Opinion Quarterly 48 (autumn 1984): 629-38. 14. John P. Robinson and Mark R. Levy, The Main Source: Learning 551 from Television News (Beverly Hills, CA: Sage, 1986). 15. C. Edward Wilson, "The Effect of Medium on Loss of Information," Journalism Quarterly 51 (spring 1974): 115; W. Russell Neuman, "Patterns of Recall Among Television News Viewers," Public Opinion Quarterly 40 (spring 1976): 115-23. 16. Doris Graber, Processing the News: How People Tame the Information Tide, 2d ed. (New York, NY: Longman, 1988). 17. Robert H. Wicks, "Improvement Over Time in Recall of Media Information: An Exploratory Study," Journal of Broadcasting & Electronic Media 36 (summer 1992): 287-302. 18. Chaffee and Kanihan, "Learning About Politics"; Berkowitz and Pritchard, "Political Knowledge." 19. Robinson and Levy, "News Media Use." 20. Zukin and Snyder, "Passive Learning." 21. Mickie Edwardson, Kurt Kent, and Maeve McConnell, "Television News Information Gain: Videotex Versus a Talking Head," Journal of Broadcasting & Electronic Media 29 (fall 1985): 367-7%) Doris Graber, "Seeing is Remembering: How Visuals Contribute to Learning from Television News," Journal of Communication 40 (summer 1990): 134-55. 22. See, for example, Alan Booth, "The Recall of News Items," Public Opinion Quarterly 34 (winter 1970-1971): 604-10. 23. Melvin L. DeFleur, Lucinda Davenport, Mary Cronin, and Margaret DeFleur, "Audience Recall of News Stories Presented by Newspaper, Computer, Television, and Radio," Journalism Quarterly 69 (winter 1992): 1010-22. 24. Christopher Harper, "The Daily Me," American Journalism Review 19 (April 1997): 41-44. 25. Thomas J. Jolmson, Mahmoud A. M. Braima, and JayanÜü Sothirajah, "Doing the Traditional Media Sidestep: Comparing the Effects of the Intemet and Other Nontraditioruil Media with Traditional Media in the 1996 Presidential Campaign," Journalism & Mass Communication Quarterly 76 (spring 1999): 99-123. 26. Andrew Kohut, "Intemet Users are on the Rise; but Public Affairs Interest Isn't," Columbia Journalism Review 38 Qanuary/February 2000): 68^9. 27. Elihu Katz, Jay G. Blumler, and Micheal Gurevitch, "Utilization of Mass Communication by the Individual," in The Uses of Mass Communications: Current Perspectives on Gratifications Research, ed. Jay G. Blumler and Elihu Katz (Beverly Hills: Sage, 1974), 19-32. 28. Douglas A. Ferguson and Elizabeth M. Perse, "The World Wide Web as a Fimctional Alternative to Television," Journal cfBroadcastingand Electronic Media 44 (spring 2000): 155-74; Andrew Flanigan and Miriam J. Metzger, "Intemet Use in the Contemporary Media Environment," HMmflKCow?MM«ica//onRcsearc/i27(January2001):153-81;ZiziPapacharissi and Alan M. Rubin, "Predictors of Intemet Use," Journal of Broadcasting and Electronic Media 44 (spring 2000): 175-96. 29. Papacharissi and Rubin, "Predictors of Intemet Use." It is important to note that Intemet use is not necessarily the same as World Wide Web use. Web content, after all, is merely a subset of Intemet traffic. 552 JouRíMLBM if MASS CCMMUMOOION QuAiemx 30. See, for example, Kohut, "Intemet Users are on the Rise." 31. "Study Reveals Web as Loosely Woven," New York Times, 18 May 2(K)0,sec.G,p. 8,col. 1. 32. David Tewksbury and Scott L. Althaus, "Differences in Knowledge Acquisition Among Readers of the Paper and Online Versions of a National Newspaper," Journalism & Mass Communication Quarterly 77 (autumn 2000): 457-79. 33. Kohut, "Intemet Users are on the Rise." 34. Ferguson and Perse, "The World Wide Web as a Functional Alternative." 35. Herbert E. Krugman, "The Impact of Television Advertising: Learning Without Involvement," Public Opinion Quarterly 29 (autumn 1965): 349-56. 36. Zukin and Snyder, "Passive Learning." 37. Grat>er, Processing the News. 38. Matthew Broersma, "Net Giants Battle for Portal Dominance," ZDNN (Online): available at http://www.zdnet.com/zdrm/stories/ news, accessed July 1999; Matthew Broersma, "You've Got E-Mail! Portals Himgry for Piece of Growing E-Commerce Pie," ZDNN (Online): available at http://www.zdnet.com/zdnn/stories/news/ 0,4586,2163823,00.html; accessed July 1999. 39. Broersma, "Net Giants Battle.' 40. Nielson//Net Ratings, "Monthly Top Ten Properties" (Online): Available at http://209.249.142.27/nnpm/owa/ Nrpublicreports.toppropertiesmonthly, accessed May 2001. 41. Nielson//Net Ratings, "Monthly Top Ten, Properties"; Media Metrix, "Top 10 Stickiest sites in November 2000; Average Minutes Spent Per Usage Month," (Online): available at h t t p : / / us.mediametrix.com/data/metrixcentral.jsp; accessed May 2001. 42. Graber, Processing the News; Wicks, "Improvement Over Time." 43. Scott L. Althaus and David Tewksbury, "Patterns of Intemet and Traditional News Media Use in a Networked Conunimity," Political Communication 17 (January-March 2000): 21-45. 44. Given that the Intemet and the World Wide Web are not entirely synonymous, it is important to note that the Pew Center surveys generally asked respondents about things they do when they "go online" (see Appendix for question wording). It is unlikely that many Americans are aware of the differences between using the Intemet and the Web, so the Pew Center may have used "going online" as a simple umbrella term. Because of popular ambiguity of the terminology, we believe there is an adequate, though not perfect, fit between Web use patterns and the Pew Center measures. 45. Robinson and Lev^, "News Media Use." 46. The Pew Center provides a demographic weighting variable with each of its data sets. Such a variable can be useful for descriptive analyses. However, application of the weight to the data dramatically increases the apparent size of each sample, which can seriously affect the significance levels of regression coefficients. Consequently, Û\e weighting factor will not be applied in the analyses reported here. 47. In the 1996 survey, the news gathering question is the same item AcoDO/awr baautieD 553 described above as online news use. 48. For nominal-by-interval analyses of this sort, eta is equivalent to a Pearson product moment correlation. Although etas do not carry a significance level, a correlation of this magnitude for a sample of this size would be significant at the .001 level. 49. The self-report item was the only incidental exposure measure appropriate for these bivariate analyses. Overall online use frequency should be a measure of incidental exposure only after the removal of variance shared with online news seeking. Multivariate tests will wait for the analyses that follow. 50. Pew Research Center, The Intemet News Audience. 51. Broersma, "Net Giants Battle." 52. America Online, the most frequently visited Web site on the Intemet (Nielsen/ /NetRatings, "Monthly Top Ten Properties"), did not post news headlines to its home page imtil after the 1996 general election John W. Mashek, Lawrence T. McGill, and Adam Clayton PoweU m. Lethargy '96: How the Media Covered a Listless Campaign [Arlington, VA: The Freedom Forum, 1997]). 53. Pew Research Center for the People and the Press, Investors Now Go Online for Quotes, Advice; Intemet Sapping Broadcast News Audience (Washington, DC: Author, 2000). 54. Vincent Price and John Zaller, "Who Gets the News? Alternative Measures of News Reception and Their Implications for Research," Public Opinion Quarterly 57 (summer 1993): 133-64. 55. Bemard Berelson, "Democratic Theory and Public Opinion," Public OpinionjQuarterly 16 (autumn 1952): 313-30. 554 Copyright of Journalism & Mass Communication Quarterly is the property of Association for Education in Journalism & Mass Communication and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.